So I'm working on this "little" project where I'm backing up YouTube videos that I'd miss if the internet were to implode (currently sitting at over 800 videos), and part of this is me going through and typing in the titles of the videos (because I didn't think ahead and store that info when made the script to download them), and by golly are there a lot of grammatical errors in some of these titles.
Also here's a lil peek of what it looks like so far...
It's got tags that can filter stuff in/out, so in this example if you want to see Jerma event streams but maybe you *really* don't like Ster, you can filter out videos that I've tagged as having him in them.
The recommendations thing stores the last 100 things you've watched and all the tags that they had. For each of those recently watched videos, it finds all videos that have three tags in common. Once it has all the similarly tagged videos for each of the previously watched ones, it randomly picks ten and lists them in the recommended videos section.
If autoplay is on, no video is selected, and the player isn't interacted with, the top one is played ten seconds after the current video finishes.
Bias towards certain creators is just because those are the ones I've added tags for so far. For instance, LGR or Michael MJD videos aren't gonna get recommended unless you've already watched one of their videos because they've not yet received tags like "long" and "informational" that a bunch of other videos would share with them, so they'd be going off their names alone.
That and the underscores in some of the video titles is because I haven't finished going through and cleaning up the json.
Melon